datascientistassoc

DragGAN is an AI-powered image editor that allows users to manipulate pictures by clicking and dragging

[{"selector":"#anim-6d18b3fc-5117-4358-8860-12e2d402b02a [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-9.23024498421834%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

It was developed by researchers from the Max Planck Institute for Informatics, MIT, and Google

[{"selector":"#anim-9118c77a-f16b-4618-a9c5-1cf7aff95ad5 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(18.87637848761366%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

DragGAN is a point-based image manipulation tool that can incrementally move multiple points of an image along a target trajectory defined by the user

[{"selector":"#anim-d1f4725a-7e29-4f56-a50a-a82fab48e0b2 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(21.67623657724911%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

The tool uses AI to keep the output within the bounds of realism

[{"selector":"#anim-730f377d-53d4-4a95-8bf9-c9eaa3f02d1d [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-11.681222474906724%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

DragGAN comprises two main components: feature-based motion supervision for handle point movement towards the target position and point tracking using specialized features to monitor the moved point.

[{"selector":"#anim-0385e56c-9319-461b-8c3f-efd7b1017960 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(17.98418952904946%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

The tool is still only a research white paper, but it has generated significant interest

[{"selector":"#anim-04bcf1e8-ea19-4336-bf1f-4ded42b99e63 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-10.623484405461385%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

Existing approaches to controlling generative adversarial networks (GANs) often lack flexibility, precision, and generality, but DragGAN offers a much less explored way of controlling GANs

[{"selector":"#anim-7a50f3e9-34e8-43d6-9335-0f61c3c08d8c [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(7.812499744007886%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

DragGAN excels at precise image deformation, making complex edits as simple as drag and drop

[{"selector":"#anim-06e697df-07f1-44d9-ad03-c1910aac3420 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-17.958543451654442%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

While DragGAN is impressive, Photoshop's comprehensive tools for graphic design still hold value

[{"selector":"#anim-1fe9f12c-89ee-422e-bd0a-4cc3212f334b [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(27.902750947125703%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

DragGAN could be the next area conquered by AI in photo editing

[{"selector":"#anim-a58c6099-15d0-42af-bf05-e3010e60cd38 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-25.430828948755135%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]